Automatic Detection and Identification of Particles with SEM/EDX

L. Görlitz, F. A. Hamrecht (IWR), M. Staudacher (Robert Bosch GmbH)

Aim

The firest major aim is to automatically, i. e. without a user set threshold on gray values, detect particles in SEM-BSD-Images. Secondly we want to use shape, texture (from SEM-Images) and spatial distribution of matter (spatially resolved EDX-spectra) of particles to identify the originating process.

Methods

The classical way to detect particles in BSD-Images is, that a threshold on the gray values is set either by a user or by standard algorithms, which expect special histogram shapes. But the grayvalue is unsuitable to properly solve this detection problem.

A promising, newly developed approach to this problem is to use morphology to describe the texture in the surrounding of a pixel (with so called morphological spectra) and then use this description to detect particles.

Fig: Comparison of classical and new approach to detecting particles